With the conventional still-picture X-ray technology currently in use, the so-called film-based technique is the most prevalent. In this technique, the patient is exposed to X-rays and the X-rays that pass through the body are then exposed onto a sheet of film. The film has the function of displaying and recording information, and is widely used throughout the world due to its capacity to be enlarged, its high degree of gradation, its light weight and ease of handling. On the other hand, the technique suffers from several disadvantages, including a complicated process of developing the image, the problem of long-term storage, and the time and effort involved in manual search and retrieval of the physical images.
By contrast, moving image photographic systems rely mainly on image intensifier (I.I). Since I.I uses the photo-electron multiplier effect inside the device, it generally has good sensitivity and has the additional advantage of exposing the patient to lower levels of radiation. The I.I technique not only provides the physician with a see-through image of the patient but also, due to the conversion of the CCD analog output to digital output (a process referred to here as digitization), makes possible the computerized recording, display and storage of such data. However, because medical diagnosis requires a high degree of gradation, even with I.I, film is often used for still picture imaging.
Recently, with a growing need to digitize X-ray images inside the hospital itself, in place of film, X-ray imaging devices that use an X-ray sensor with solid-state image sensing elements arrayed two-dimensionally to convert the X-ray image into electrical signals have begun to be used. Since the X-ray image can then be replaced with digital information, image information can be sent instantaneously to distant locations, with the advantage of being able to provide state-of-the-art, high-quality diagnostics even to remote areas. Moreover, if no film is used the space previously required for its storage can be turned to other, more productive uses. If in the future it becomes possible to introduce more advanced and sophisticated image processing techniques, it is possible that diagnostics may to some extent be computerized and therefore automated, without the intervention of a radiologist.
Moreover, in recent years, with the use of amorphous thin-film semiconductors in solid-state image sensing elements, X-ray imaging devices capable of taking still pictures have been developed. Using amorphous silicon thin-film semiconductor production technology, photos exceeding 40 cm a side and capable of completely imaging the human torso have been commercialized. Since the production process itself is relatively simple, it is expected that inexpensive detectors based on this technique will become available in the not-so-distant future. In addition, since amorphous silicon can be produced in thin glass sheets having a thickness of 1 mm or less, the detector itself can be made very thin and compact, for greater ease of handling.
An X-ray imaging device of the sort described above typically has a photoelectric converter circuit, on which a plurality of photoelectric converters for converting radiation into electrical signals are arranged as in a matrix, and a read-out circuit for reading out the electrical signals form the photoelectric converter circuit. FIG. 7, for example, is a timing chart showing the operation of an X-ray imaging device having nine photoelectric converters S1-1 through S3-3 (including switches (TFT) T1-1-T3-3). As for the photoelectric conversion interval (given in the diagram as the X-ray exposure interval): In a state in which the TFT are all OFF, when the light source (X-rays) are turned ON in pulses, the respective photoelectric converters are struck by the light and a signal electric charge comparable to the amount of light is stored in the converter capacitor. If a fluorescent material is used to convert the X-rays into visible light, then either a light-guiding member for guiding the light made visible in proportion to the number of X-rays to the photoelectric converters may be used or the fluorescent material may be disposed near the electrodes of the photoelectric converters.
It should be noted that the signal electrical charge is held in the converter capacitor after the light source is OFF.
Next, as for the read-out interval: The read-out is accomplished at the S1-1-S3-3, one row at a time, starting with row S1-1-S1-3, then with row S2-1-S2-3, and finally with row S3-1-S3-3. First, a gate pulse is applied from SR1 to the switches (TFT) T1-1-T1-3 gate lines in order to read out the first row S1-1-S1-3. Doing so turns T1-1-T1-3 ON and the signal electrical charge that had been stored in S1-1-S1-3 is sent to the signal lines M1-M3 to which read-out capacitors CM1-CM3 have been added, so that the signal electrical charge is sent to the capacitors CM1-CM3 via the TFT. For example, read-out capacitor CM1 added to signal line M1 is the (three-)sum total of the T1-1-T1-3 gate-source interelectrode capacitance (Cgs). Amps A1-A3 amplify the signal electrical charge sent to signal lines M1-M3.
The amplified signal electrical charge sent to capacitors CL1-CL3 both turns OFF and holds SMPL signal OFF. Next, by imparting a pulse from a shift resister 103 to switches Sr1, Sr2 and Sr3 (in that order) the signals held at CL1-CL3 are then output from an amp 104 in the order CL1, CL2 and CL3. Since analog signal outputs B1, B2 and B3 are output from the amp 104, the entire unit, including the shift resister 103 and the switches Sr1-Sr3, is called an analog multiplexer. Ultimately, one row's worth of photoelectric conversion signals (S1-1, S1-2, S1-3) is output in sequence by the analog multiplexer. The read-out of the second row S2-1-S2-3 is carried out in the same way as the read-out of the first row described above.
If the signals at signals lines M1-M3 are sampled and held at CL1-CL3 by the first row's SMPL signal, then the signal lines M1-M3 can be reset to ground electric potential by a CRES signal and thereafter a G2 gate pulse can be applied. In other words, second-row signal electrical charges from the photoelectric converters S2-1-S2-3 can be transmitted by the SR1 while at the same time the first row's signals are being serially converted by the SR2. In so doing, all the signal electrical charges of the first through third rows of photoelectric converters can be output.
The photoelectric converter circuit operation described above is capable of reading X-ray images. But in reality the image as such also contains fixed pattern noise generated by the photoelectric converter circuit and the read-out circuit.
There are a number of factors that cause fixed pattern noise, including (A) variations in photoelectric converter dark current, (B) variations in switch leak current, (C) variations in photoelectric converter circuit signal wire capacitors CM1-CM3 and (D) variations in read-out circuit amp (for example A1-A3) offset voltage.
Whereas causes (A) and (B) described above show up as dots in the final image, causes (C) and (D) appear as lines; in each case, they degrade the quality of the image. Therefore, conventionally, this type of fixed pattern noise is read out as a dark output image, which is then subtracted from the X-ray image so as to correct the fixed pattern noise (hereinafter FPN).
The dark output image itself acquired in one of two ways: it is either read in when the unit is shipped from the factory or prior to photographing and then stored in memory as dark output data, or it is obtained just before or just after read-in of an X-ray image.
However, there are problems with the conventional methods of acquiring the dark output image (FPN).
In the former case, when the FPN is set either at shipment or prior to photography, there are two problems:
(a) The dark currents of the photoelectric converters have temperature characteristics, which may be different when the FPN is obtained and when the X-ray image is shot. Such differences in temperature characteristics can prevent the FPN from being fully corrected.
(b) Since the TFT leak current changes over time, the leak current may be different when the FPN is obtained and when the X-ray image is shot, and such differences in leak current can prevent the FPN from being fully corrected.
The technique of reading in the FPN just before or just after taking the X-ray image, although it solves the two problems described above, has a problem of its own:
(c) random noise generated by the photoelectric converter circuit (hereinafter referred to simply as random noise) as well as random noise generated by the read-out circuit, which is not fixed pattern noise and therefore cannot be fully corrected, and which, when processed for elimination, only increases by a factor of √{square root over ( )}2 and degrades the quality of the image.
Random noise is caused by shot noise from the photoelectric converter dark electrical charge and by heat noise (that is, Johnson noise) caused by electron thermal disturbances inside the switches. It is one type of noise that is unavoidable. In addition, if the signal lines M1-M3 have any appreciable internal resistance, that resistance, too, will generate Johnson noise or random noise. The operational amplifier that forms part of the read-out circuit also generates random noise.
The amount or volume of random noise can be determined by focusing on a particular pixel and sampling its dark data several times in order to obtain its dark data distribution. In other words, that distribution is a Gaussian distribution, i.e., noise that has a frequency distribution that follows the Gaussian curve, and can be calculated by obtaining the standard deviation σ.
The dark image output from an X-ray imaging apparatus comprising the photoelectric converter circuit in which the photoelectric converter are arranged in a two-dimensional array, contains both FPN as well as random noise. Moreover, both types of noise appear not only in the dark image output but also in the X-ray image as well. Subtracting the dark image output from the X-ray image cancels the FPN but leaves the random noise uncorrected. Also, between the dark image output and the X-ray image, the random noise itself can differ from one pixel to the next. The random noise generated when reading in the dark image output and the random noise generated when reading in the X-ray image are mutually independent phenomena, so an image in which the former is subtracted from the latter (that is, FPN correction) will have a volume of random noise (√{square root over ( )}2·σ) that is √{square root over ( )}2 the volume of random noise before correction (standard deviation σ).